dir.c 24.8 KB
Newer Older
M
Miklos Szeredi 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/*
 *
 * Copyright (C) 2011 Novell Inc.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 as published by
 * the Free Software Foundation.
 */

#include <linux/fs.h>
#include <linux/namei.h>
#include <linux/xattr.h>
#include <linux/security.h>
#include <linux/cred.h>
15
#include <linux/module.h>
16 17
#include <linux/posix_acl.h>
#include <linux/posix_acl_xattr.h>
18
#include <linux/atomic.h>
M
Miklos Szeredi 已提交
19
#include <linux/ratelimit.h>
M
Miklos Szeredi 已提交
20 21
#include "overlayfs.h"

22 23 24 25 26
static unsigned short ovl_redirect_max = 256;
module_param_named(redirect_max, ovl_redirect_max, ushort, 0644);
MODULE_PARM_DESC(ovl_redirect_max,
		 "Maximum length of absolute redirect xattr value");

27
int ovl_cleanup(struct inode *wdir, struct dentry *wdentry)
M
Miklos Szeredi 已提交
28 29 30 31
{
	int err;

	dget(wdentry);
32
	if (d_is_dir(wdentry))
M
Miklos Szeredi 已提交
33 34 35 36 37 38 39 40 41
		err = ovl_do_rmdir(wdir, wdentry);
	else
		err = ovl_do_unlink(wdir, wdentry);
	dput(wdentry);

	if (err) {
		pr_err("overlayfs: cleanup of '%pd2' failed (%i)\n",
		       wdentry, err);
	}
42 43

	return err;
M
Miklos Szeredi 已提交
44 45
}

46
struct dentry *ovl_lookup_temp(struct dentry *workdir)
M
Miklos Szeredi 已提交
47 48 49
{
	struct dentry *temp;
	char name[20];
50
	static atomic_t temp_id = ATOMIC_INIT(0);
M
Miklos Szeredi 已提交
51

52 53
	/* counter is allowed to wrap, since temp dentries are ephemeral */
	snprintf(name, sizeof(name), "#%x", atomic_inc_return(&temp_id));
M
Miklos Szeredi 已提交
54 55 56 57 58 59 60 61 62 63 64 65

	temp = lookup_one_len(name, workdir, strlen(name));
	if (!IS_ERR(temp) && temp->d_inode) {
		pr_err("overlayfs: workdir/%s already exists\n", name);
		dput(temp);
		temp = ERR_PTR(-EIO);
	}

	return temp;
}

/* caller holds i_mutex on workdir */
66
static struct dentry *ovl_whiteout(struct dentry *workdir)
M
Miklos Szeredi 已提交
67 68 69 70 71
{
	int err;
	struct dentry *whiteout;
	struct inode *wdir = workdir->d_inode;

72
	whiteout = ovl_lookup_temp(workdir);
M
Miklos Szeredi 已提交
73 74 75 76 77 78 79 80 81 82 83 84
	if (IS_ERR(whiteout))
		return whiteout;

	err = ovl_do_whiteout(wdir, whiteout);
	if (err) {
		dput(whiteout);
		whiteout = ERR_PTR(err);
	}

	return whiteout;
}

85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116
/* Caller must hold i_mutex on both workdir and dir */
int ovl_cleanup_and_whiteout(struct dentry *workdir, struct inode *dir,
			     struct dentry *dentry)
{
	struct inode *wdir = workdir->d_inode;
	struct dentry *whiteout;
	int err;
	int flags = 0;

	whiteout = ovl_whiteout(workdir);
	err = PTR_ERR(whiteout);
	if (IS_ERR(whiteout))
		return err;

	if (d_is_dir(dentry))
		flags = RENAME_EXCHANGE;

	err = ovl_do_rename(wdir, whiteout, dir, dentry, flags);
	if (err)
		goto kill_whiteout;
	if (flags)
		ovl_cleanup(wdir, dentry);

out:
	dput(whiteout);
	return err;

kill_whiteout:
	ovl_cleanup(wdir, whiteout);
	goto out;
}

M
Miklos Szeredi 已提交
117
int ovl_create_real(struct inode *dir, struct dentry *newdentry,
A
Al Viro 已提交
118
		    struct cattr *attr, struct dentry *hardlink, bool debug)
M
Miklos Szeredi 已提交
119 120 121 122 123 124 125 126 127
{
	int err;

	if (newdentry->d_inode)
		return -ESTALE;

	if (hardlink) {
		err = ovl_do_link(hardlink, dir, newdentry, debug);
	} else {
A
Al Viro 已提交
128
		switch (attr->mode & S_IFMT) {
M
Miklos Szeredi 已提交
129
		case S_IFREG:
A
Al Viro 已提交
130
			err = ovl_do_create(dir, newdentry, attr->mode, debug);
M
Miklos Szeredi 已提交
131 132 133
			break;

		case S_IFDIR:
A
Al Viro 已提交
134
			err = ovl_do_mkdir(dir, newdentry, attr->mode, debug);
M
Miklos Szeredi 已提交
135 136 137 138 139 140 141
			break;

		case S_IFCHR:
		case S_IFBLK:
		case S_IFIFO:
		case S_IFSOCK:
			err = ovl_do_mknod(dir, newdentry,
A
Al Viro 已提交
142
					   attr->mode, attr->rdev, debug);
M
Miklos Szeredi 已提交
143 144 145
			break;

		case S_IFLNK:
A
Al Viro 已提交
146
			err = ovl_do_symlink(dir, newdentry, attr->link, debug);
M
Miklos Szeredi 已提交
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162
			break;

		default:
			err = -EPERM;
		}
	}
	if (!err && WARN_ON(!newdentry->d_inode)) {
		/*
		 * Not quite sure if non-instantiated dentry is legal or not.
		 * VFS doesn't seem to care so check and warn here.
		 */
		err = -ENOENT;
	}
	return err;
}

163 164
static int ovl_set_opaque_xerr(struct dentry *dentry, struct dentry *upper,
			       int xerr)
M
Miklos Szeredi 已提交
165
{
M
Miklos Szeredi 已提交
166 167
	int err;

168
	err = ovl_check_setxattr(dentry, upper, OVL_XATTR_OPAQUE, "y", 1, xerr);
M
Miklos Szeredi 已提交
169 170 171 172
	if (!err)
		ovl_dentry_set_opaque(dentry);

	return err;
M
Miklos Szeredi 已提交
173 174
}

175 176 177 178 179 180 181 182 183 184
static int ovl_set_opaque(struct dentry *dentry, struct dentry *upperdentry)
{
	/*
	 * Fail with -EIO when trying to create opaque dir and upper doesn't
	 * support xattrs. ovl_rename() calls ovl_set_opaque_xerr(-EXDEV) to
	 * return a specific error for noxattr case.
	 */
	return ovl_set_opaque_xerr(dentry, upperdentry, -EIO);
}

185 186
/* Common operations required to be done after creation of file on upper */
static void ovl_instantiate(struct dentry *dentry, struct inode *inode,
M
Miklos Szeredi 已提交
187
			    struct dentry *newdentry, bool hardlink)
188
{
189
	ovl_dentry_version_inc(dentry->d_parent, false);
190
	ovl_dentry_set_upper_alias(dentry);
M
Miklos Szeredi 已提交
191
	if (!hardlink) {
192
		ovl_inode_update(inode, newdentry);
M
Miklos Szeredi 已提交
193 194
		ovl_copyattr(newdentry->d_inode, inode);
	} else {
195 196
		WARN_ON(ovl_inode_real(inode) != d_inode(newdentry));
		dput(newdentry);
M
Miklos Szeredi 已提交
197 198
		inc_nlink(inode);
	}
199
	d_instantiate(dentry, inode);
200 201 202
	/* Force lookup of new upper hardlink to find its lower */
	if (hardlink)
		d_drop(dentry);
203 204
}

205 206 207 208 209
static bool ovl_type_merge(struct dentry *dentry)
{
	return OVL_TYPE_MERGE(ovl_path_type(dentry));
}

210 211 212 213 214
static bool ovl_type_origin(struct dentry *dentry)
{
	return OVL_TYPE_ORIGIN(ovl_path_type(dentry));
}

M
Miklos Szeredi 已提交
215
static int ovl_create_upper(struct dentry *dentry, struct inode *inode,
A
Al Viro 已提交
216
			    struct cattr *attr, struct dentry *hardlink)
M
Miklos Szeredi 已提交
217 218 219 220 221 222
{
	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
	struct inode *udir = upperdir->d_inode;
	struct dentry *newdentry;
	int err;

223
	if (!hardlink && !IS_POSIXACL(udir))
A
Al Viro 已提交
224
		attr->mode &= ~current_umask();
225

A
Al Viro 已提交
226
	inode_lock_nested(udir, I_MUTEX_PARENT);
M
Miklos Szeredi 已提交
227 228 229 230 231
	newdentry = lookup_one_len(dentry->d_name.name, upperdir,
				   dentry->d_name.len);
	err = PTR_ERR(newdentry);
	if (IS_ERR(newdentry))
		goto out_unlock;
A
Al Viro 已提交
232
	err = ovl_create_real(udir, newdentry, attr, hardlink, false);
M
Miklos Szeredi 已提交
233 234 235
	if (err)
		goto out_dput;

236
	if (ovl_type_merge(dentry->d_parent) && d_is_dir(newdentry)) {
237 238 239 240
		/* Setting opaque here is just an optimization, allow to fail */
		ovl_set_opaque(dentry, newdentry);
	}

M
Miklos Szeredi 已提交
241
	ovl_instantiate(dentry, inode, newdentry, !!hardlink);
M
Miklos Szeredi 已提交
242 243 244 245
	newdentry = NULL;
out_dput:
	dput(newdentry);
out_unlock:
A
Al Viro 已提交
246
	inode_unlock(udir);
M
Miklos Szeredi 已提交
247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262
	return err;
}

static struct dentry *ovl_clear_empty(struct dentry *dentry,
				      struct list_head *list)
{
	struct dentry *workdir = ovl_workdir(dentry);
	struct inode *wdir = workdir->d_inode;
	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
	struct inode *udir = upperdir->d_inode;
	struct path upperpath;
	struct dentry *upper;
	struct dentry *opaquedir;
	struct kstat stat;
	int err;

263 264 265
	if (WARN_ON(!workdir))
		return ERR_PTR(-EROFS);

M
Miklos Szeredi 已提交
266 267 268 269 270
	err = ovl_lock_rename_workdir(workdir, upperdir);
	if (err)
		goto out;

	ovl_path_upper(dentry, &upperpath);
271 272
	err = vfs_getattr(&upperpath, &stat,
			  STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
M
Miklos Szeredi 已提交
273 274 275 276 277 278 279 280 281 282
	if (err)
		goto out_unlock;

	err = -ESTALE;
	if (!S_ISDIR(stat.mode))
		goto out_unlock;
	upper = upperpath.dentry;
	if (upper->d_parent->d_inode != udir)
		goto out_unlock;

283
	opaquedir = ovl_lookup_temp(workdir);
M
Miklos Szeredi 已提交
284 285 286 287
	err = PTR_ERR(opaquedir);
	if (IS_ERR(opaquedir))
		goto out_unlock;

A
Al Viro 已提交
288 289
	err = ovl_create_real(wdir, opaquedir,
			      &(struct cattr){.mode = stat.mode}, NULL, true);
M
Miklos Szeredi 已提交
290 291 292 293 294 295 296
	if (err)
		goto out_dput;

	err = ovl_copy_xattr(upper, opaquedir);
	if (err)
		goto out_cleanup;

M
Miklos Szeredi 已提交
297
	err = ovl_set_opaque(dentry, opaquedir);
M
Miklos Szeredi 已提交
298 299 300
	if (err)
		goto out_cleanup;

A
Al Viro 已提交
301
	inode_lock(opaquedir->d_inode);
M
Miklos Szeredi 已提交
302
	err = ovl_set_attr(opaquedir, &stat);
A
Al Viro 已提交
303
	inode_unlock(opaquedir->d_inode);
M
Miklos Szeredi 已提交
304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329
	if (err)
		goto out_cleanup;

	err = ovl_do_rename(wdir, opaquedir, udir, upper, RENAME_EXCHANGE);
	if (err)
		goto out_cleanup;

	ovl_cleanup_whiteouts(upper, list);
	ovl_cleanup(wdir, upper);
	unlock_rename(workdir, upperdir);

	/* dentry's upper doesn't match now, get rid of it */
	d_drop(dentry);

	return opaquedir;

out_cleanup:
	ovl_cleanup(wdir, opaquedir);
out_dput:
	dput(opaquedir);
out_unlock:
	unlock_rename(workdir, upperdir);
out:
	return ERR_PTR(err);
}

330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
static int ovl_set_upper_acl(struct dentry *upperdentry, const char *name,
			     const struct posix_acl *acl)
{
	void *buffer;
	size_t size;
	int err;

	if (!IS_ENABLED(CONFIG_FS_POSIX_ACL) || !acl)
		return 0;

	size = posix_acl_to_xattr(NULL, acl, NULL, 0);
	buffer = kmalloc(size, GFP_KERNEL);
	if (!buffer)
		return -ENOMEM;

	size = posix_acl_to_xattr(&init_user_ns, acl, buffer, size);
	err = size;
	if (err < 0)
		goto out_free;

	err = vfs_setxattr(upperdentry, name, buffer, size, XATTR_CREATE);
out_free:
	kfree(buffer);
	return err;
}

M
Miklos Szeredi 已提交
356
static int ovl_create_over_whiteout(struct dentry *dentry, struct inode *inode,
A
Al Viro 已提交
357
				    struct cattr *cattr,
M
Miklos Szeredi 已提交
358 359 360 361 362 363 364 365 366
				    struct dentry *hardlink)
{
	struct dentry *workdir = ovl_workdir(dentry);
	struct inode *wdir = workdir->d_inode;
	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
	struct inode *udir = upperdir->d_inode;
	struct dentry *upper;
	struct dentry *newdentry;
	int err;
367
	struct posix_acl *acl, *default_acl;
M
Miklos Szeredi 已提交
368

369 370 371
	if (WARN_ON(!workdir))
		return -EROFS;

372 373
	if (!hardlink) {
		err = posix_acl_create(dentry->d_parent->d_inode,
A
Al Viro 已提交
374
				       &cattr->mode, &default_acl, &acl);
375 376 377 378
		if (err)
			return err;
	}

M
Miklos Szeredi 已提交
379 380 381 382
	err = ovl_lock_rename_workdir(workdir, upperdir);
	if (err)
		goto out;

383
	newdentry = ovl_lookup_temp(workdir);
M
Miklos Szeredi 已提交
384 385 386 387 388 389 390 391 392 393
	err = PTR_ERR(newdentry);
	if (IS_ERR(newdentry))
		goto out_unlock;

	upper = lookup_one_len(dentry->d_name.name, upperdir,
			       dentry->d_name.len);
	err = PTR_ERR(upper);
	if (IS_ERR(upper))
		goto out_dput;

A
Al Viro 已提交
394
	err = ovl_create_real(wdir, newdentry, cattr, hardlink, true);
M
Miklos Szeredi 已提交
395 396 397
	if (err)
		goto out_dput2;

M
Miklos Szeredi 已提交
398 399 400
	/*
	 * mode could have been mutilated due to umask (e.g. sgid directory)
	 */
M
Miklos Szeredi 已提交
401
	if (!hardlink &&
A
Al Viro 已提交
402 403
	    !S_ISLNK(cattr->mode) &&
	    newdentry->d_inode->i_mode != cattr->mode) {
M
Miklos Szeredi 已提交
404 405
		struct iattr attr = {
			.ia_valid = ATTR_MODE,
A
Al Viro 已提交
406
			.ia_mode = cattr->mode,
M
Miklos Szeredi 已提交
407 408 409 410 411 412 413
		};
		inode_lock(newdentry->d_inode);
		err = notify_change(newdentry, &attr, NULL);
		inode_unlock(newdentry->d_inode);
		if (err)
			goto out_cleanup;
	}
414 415 416 417 418 419 420 421 422 423 424
	if (!hardlink) {
		err = ovl_set_upper_acl(newdentry, XATTR_NAME_POSIX_ACL_ACCESS,
					acl);
		if (err)
			goto out_cleanup;

		err = ovl_set_upper_acl(newdentry, XATTR_NAME_POSIX_ACL_DEFAULT,
					default_acl);
		if (err)
			goto out_cleanup;
	}
M
Miklos Szeredi 已提交
425

A
Al Viro 已提交
426
	if (!hardlink && S_ISDIR(cattr->mode)) {
M
Miklos Szeredi 已提交
427
		err = ovl_set_opaque(dentry, newdentry);
M
Miklos Szeredi 已提交
428 429 430 431 432 433 434 435 436 437 438 439 440 441
		if (err)
			goto out_cleanup;

		err = ovl_do_rename(wdir, newdentry, udir, upper,
				    RENAME_EXCHANGE);
		if (err)
			goto out_cleanup;

		ovl_cleanup(wdir, upper);
	} else {
		err = ovl_do_rename(wdir, newdentry, udir, upper, 0);
		if (err)
			goto out_cleanup;
	}
M
Miklos Szeredi 已提交
442
	ovl_instantiate(dentry, inode, newdentry, !!hardlink);
M
Miklos Szeredi 已提交
443 444 445 446 447 448 449 450
	newdentry = NULL;
out_dput2:
	dput(upper);
out_dput:
	dput(newdentry);
out_unlock:
	unlock_rename(workdir, upperdir);
out:
451 452 453 454
	if (!hardlink) {
		posix_acl_release(acl);
		posix_acl_release(default_acl);
	}
M
Miklos Szeredi 已提交
455 456 457 458 459 460 461
	return err;

out_cleanup:
	ovl_cleanup(wdir, newdentry);
	goto out_dput2;
}

M
Miklos Szeredi 已提交
462
static int ovl_create_or_link(struct dentry *dentry, struct inode *inode,
463 464
			      struct cattr *attr, struct dentry *hardlink,
			      bool origin)
M
Miklos Szeredi 已提交
465 466
{
	int err;
467 468
	const struct cred *old_cred;
	struct cred *override_cred;
469
	struct dentry *parent = dentry->d_parent;
M
Miklos Szeredi 已提交
470

471
	err = ovl_copy_up(parent);
M
Miklos Szeredi 已提交
472
	if (err)
M
Miklos Szeredi 已提交
473
		return err;
M
Miklos Szeredi 已提交
474

475
	old_cred = ovl_override_creds(dentry->d_sb);
476 477 478 479 480 481 482 483 484 485 486

	/*
	 * When linking a file with copy up origin into a new parent, mark the
	 * new parent dir "impure".
	 */
	if (origin) {
		err = ovl_set_impure(parent, ovl_dentry_upper(parent));
		if (err)
			goto out_revert_creds;
	}

487 488 489
	err = -ENOMEM;
	override_cred = prepare_creds();
	if (override_cred) {
M
Miklos Szeredi 已提交
490 491
		override_cred->fsuid = inode->i_uid;
		override_cred->fsgid = inode->i_gid;
492 493
		if (!hardlink) {
			err = security_dentry_create_files_as(dentry,
A
Al Viro 已提交
494
					attr->mode, &dentry->d_name, old_cred,
495 496 497 498 499 500
					override_cred);
			if (err) {
				put_cred(override_cred);
				goto out_revert_creds;
			}
		}
501 502 503
		put_cred(override_creds(override_cred));
		put_cred(override_cred);

M
Miklos Szeredi 已提交
504
		if (!ovl_dentry_is_whiteout(dentry))
A
Al Viro 已提交
505
			err = ovl_create_upper(dentry, inode, attr,
506 507
						hardlink);
		else
A
Al Viro 已提交
508 509
			err = ovl_create_over_whiteout(dentry, inode, attr,
							hardlink);
M
Miklos Szeredi 已提交
510
	}
511
out_revert_creds:
512
	revert_creds(old_cred);
M
Miklos Szeredi 已提交
513 514 515 516 517 518 519
	if (!err) {
		struct inode *realinode = d_inode(ovl_dentry_upper(dentry));

		WARN_ON(inode->i_mode != realinode->i_mode);
		WARN_ON(!uid_eq(inode->i_uid, realinode->i_uid));
		WARN_ON(!gid_eq(inode->i_gid, realinode->i_gid));
	}
M
Miklos Szeredi 已提交
520 521 522 523 524 525 526
	return err;
}

static int ovl_create_object(struct dentry *dentry, int mode, dev_t rdev,
			     const char *link)
{
	int err;
M
Miklos Szeredi 已提交
527
	struct inode *inode;
A
Al Viro 已提交
528
	struct cattr attr = {
M
Miklos Szeredi 已提交
529
		.rdev = rdev,
A
Al Viro 已提交
530
		.link = link,
M
Miklos Szeredi 已提交
531
	};
M
Miklos Szeredi 已提交
532 533

	err = ovl_want_write(dentry);
M
Miklos Szeredi 已提交
534 535 536 537
	if (err)
		goto out;

	err = -ENOMEM;
538
	inode = ovl_new_inode(dentry->d_sb, mode, rdev);
M
Miklos Szeredi 已提交
539 540 541 542
	if (!inode)
		goto out_drop_write;

	inode_init_owner(inode, dentry->d_parent->d_inode, mode);
A
Al Viro 已提交
543
	attr.mode = inode->i_mode;
M
Miklos Szeredi 已提交
544

545
	err = ovl_create_or_link(dentry, inode, &attr, NULL, false);
M
Miklos Szeredi 已提交
546 547
	if (err)
		iput(inode);
M
Miklos Szeredi 已提交
548

M
Miklos Szeredi 已提交
549 550 551
out_drop_write:
	ovl_drop_write(dentry);
out:
M
Miklos Szeredi 已提交
552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
	return err;
}

static int ovl_create(struct inode *dir, struct dentry *dentry, umode_t mode,
		      bool excl)
{
	return ovl_create_object(dentry, (mode & 07777) | S_IFREG, 0, NULL);
}

static int ovl_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
{
	return ovl_create_object(dentry, (mode & 07777) | S_IFDIR, 0, NULL);
}

static int ovl_mknod(struct inode *dir, struct dentry *dentry, umode_t mode,
		     dev_t rdev)
{
	/* Don't allow creation of "whiteout" on overlay */
	if (S_ISCHR(mode) && rdev == WHITEOUT_DEV)
		return -EPERM;

	return ovl_create_object(dentry, mode, rdev, NULL);
}

static int ovl_symlink(struct inode *dir, struct dentry *dentry,
		       const char *link)
{
	return ovl_create_object(dentry, S_IFLNK, 0, link);
}

static int ovl_link(struct dentry *old, struct inode *newdir,
		    struct dentry *new)
{
	int err;
586
	bool locked = false;
M
Miklos Szeredi 已提交
587
	struct inode *inode;
M
Miklos Szeredi 已提交
588 589 590 591 592 593 594 595 596

	err = ovl_want_write(old);
	if (err)
		goto out;

	err = ovl_copy_up(old);
	if (err)
		goto out_drop_write;

597 598 599 600
	err = ovl_nlink_start(old, &locked);
	if (err)
		goto out_drop_write;

M
Miklos Szeredi 已提交
601 602 603
	inode = d_inode(old);
	ihold(inode);

604 605
	err = ovl_create_or_link(new, inode, NULL, ovl_dentry_upper(old),
				 ovl_type_origin(old));
M
Miklos Szeredi 已提交
606 607
	if (err)
		iput(inode);
M
Miklos Szeredi 已提交
608

609
	ovl_nlink_end(old, locked);
M
Miklos Szeredi 已提交
610 611 612 613 614 615
out_drop_write:
	ovl_drop_write(old);
out:
	return err;
}

M
Miklos Szeredi 已提交
616 617 618 619 620
static bool ovl_matches_upper(struct dentry *dentry, struct dentry *upper)
{
	return d_inode(ovl_dentry_upper(dentry)) == d_inode(upper);
}

621 622
static int ovl_remove_and_whiteout(struct dentry *dentry,
				   struct list_head *list)
M
Miklos Szeredi 已提交
623 624 625 626 627 628 629
{
	struct dentry *workdir = ovl_workdir(dentry);
	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
	struct dentry *upper;
	struct dentry *opaquedir = NULL;
	int err;

630 631 632
	if (WARN_ON(!workdir))
		return -EROFS;

633 634
	if (!list_empty(list)) {
		opaquedir = ovl_clear_empty(dentry, list);
M
Miklos Szeredi 已提交
635 636 637
		err = PTR_ERR(opaquedir);
		if (IS_ERR(opaquedir))
			goto out;
M
Miklos Szeredi 已提交
638 639 640 641 642 643
	}

	err = ovl_lock_rename_workdir(workdir, upperdir);
	if (err)
		goto out_dput;

644 645 646 647
	upper = lookup_one_len(dentry->d_name.name, upperdir,
			       dentry->d_name.len);
	err = PTR_ERR(upper);
	if (IS_ERR(upper))
M
Miklos Szeredi 已提交
648 649
		goto out_unlock;

650 651 652
	err = -ESTALE;
	if ((opaquedir && upper != opaquedir) ||
	    (!opaquedir && ovl_dentry_upper(dentry) &&
M
Miklos Szeredi 已提交
653
	     !ovl_matches_upper(dentry, upper))) {
654 655
		goto out_dput_upper;
	}
M
Miklos Szeredi 已提交
656

657
	err = ovl_cleanup_and_whiteout(workdir, d_inode(upperdir), upper);
658
	if (err)
659
		goto out_d_drop;
M
Miklos Szeredi 已提交
660

661
	ovl_dentry_version_inc(dentry->d_parent, true);
M
Miklos Szeredi 已提交
662 663
out_d_drop:
	d_drop(dentry);
664 665
out_dput_upper:
	dput(upper);
M
Miklos Szeredi 已提交
666 667 668 669 670 671 672 673
out_unlock:
	unlock_rename(workdir, upperdir);
out_dput:
	dput(opaquedir);
out:
	return err;
}

674 675
static int ovl_remove_upper(struct dentry *dentry, bool is_dir,
			    struct list_head *list)
M
Miklos Szeredi 已提交
676 677 678
{
	struct dentry *upperdir = ovl_dentry_upper(dentry->d_parent);
	struct inode *dir = upperdir->d_inode;
679
	struct dentry *upper;
680
	struct dentry *opaquedir = NULL;
M
Miklos Szeredi 已提交
681 682
	int err;

683 684
	if (!list_empty(list)) {
		opaquedir = ovl_clear_empty(dentry, list);
685 686 687 688 689
		err = PTR_ERR(opaquedir);
		if (IS_ERR(opaquedir))
			goto out;
	}

A
Al Viro 已提交
690
	inode_lock_nested(dir, I_MUTEX_PARENT);
691 692 693 694 695 696
	upper = lookup_one_len(dentry->d_name.name, upperdir,
			       dentry->d_name.len);
	err = PTR_ERR(upper);
	if (IS_ERR(upper))
		goto out_unlock;

M
Miklos Szeredi 已提交
697
	err = -ESTALE;
698
	if ((opaquedir && upper != opaquedir) ||
M
Miklos Szeredi 已提交
699
	    (!opaquedir && !ovl_matches_upper(dentry, upper)))
700 701 702 703 704 705
		goto out_dput_upper;

	if (is_dir)
		err = vfs_rmdir(dir, upper);
	else
		err = vfs_unlink(dir, upper, NULL);
706
	ovl_dentry_version_inc(dentry->d_parent, ovl_type_origin(dentry));
M
Miklos Szeredi 已提交
707 708 709 710 711 712 713

	/*
	 * Keeping this dentry hashed would mean having to release
	 * upperpath/lowerpath, which could only be done if we are the
	 * sole user of this dentry.  Too tricky...  Just unhash for
	 * now.
	 */
714 715
	if (!err)
		d_drop(dentry);
716 717
out_dput_upper:
	dput(upper);
718
out_unlock:
A
Al Viro 已提交
719
	inode_unlock(dir);
720 721
	dput(opaquedir);
out:
M
Miklos Szeredi 已提交
722 723 724
	return err;
}

725 726 727 728 729 730
static bool ovl_pure_upper(struct dentry *dentry)
{
	return !ovl_dentry_lower(dentry) &&
	       !ovl_test_flag(OVL_WHITEOUTS, d_inode(dentry));
}

M
Miklos Szeredi 已提交
731 732 733
static int ovl_do_remove(struct dentry *dentry, bool is_dir)
{
	int err;
734
	bool locked = false;
735
	const struct cred *old_cred;
736 737 738 739 740 741 742 743 744
	bool lower_positive = ovl_lower_positive(dentry);
	LIST_HEAD(list);

	/* No need to clean pure upper removed by vfs_rmdir() */
	if (is_dir && (lower_positive || !ovl_pure_upper(dentry))) {
		err = ovl_check_empty_dir(dentry, &list);
		if (err)
			goto out;
	}
745

M
Miklos Szeredi 已提交
746 747 748 749 750 751 752 753
	err = ovl_want_write(dentry);
	if (err)
		goto out;

	err = ovl_copy_up(dentry->d_parent);
	if (err)
		goto out_drop_write;

754 755 756
	err = ovl_nlink_start(dentry, &locked);
	if (err)
		goto out_drop_write;
M
Miklos Szeredi 已提交
757

758
	old_cred = ovl_override_creds(dentry->d_sb);
759 760
	if (!lower_positive)
		err = ovl_remove_upper(dentry, is_dir, &list);
761
	else
762
		err = ovl_remove_and_whiteout(dentry, &list);
763
	revert_creds(old_cred);
M
Miklos Szeredi 已提交
764 765 766 767 768 769
	if (!err) {
		if (is_dir)
			clear_nlink(dentry->d_inode);
		else
			drop_nlink(dentry->d_inode);
	}
770
	ovl_nlink_end(dentry, locked);
M
Miklos Szeredi 已提交
771 772 773
out_drop_write:
	ovl_drop_write(dentry);
out:
774
	ovl_cache_free(&list);
M
Miklos Szeredi 已提交
775 776 777 778 779 780 781 782 783 784 785 786 787
	return err;
}

static int ovl_unlink(struct inode *dir, struct dentry *dentry)
{
	return ovl_do_remove(dentry, false);
}

static int ovl_rmdir(struct inode *dir, struct dentry *dentry)
{
	return ovl_do_remove(dentry, true);
}

788 789 790 791 792 793 794
static bool ovl_type_merge_or_lower(struct dentry *dentry)
{
	enum ovl_path_type type = ovl_path_type(dentry);

	return OVL_TYPE_MERGE(type) || !OVL_TYPE_UPPER(type);
}

M
Miklos Szeredi 已提交
795 796 797 798 799 800 801 802 803 804
static bool ovl_can_move(struct dentry *dentry)
{
	return ovl_redirect_dir(dentry->d_sb) ||
		!d_is_dir(dentry) || !ovl_type_merge_or_lower(dentry);
}

static char *ovl_get_redirect(struct dentry *dentry, bool samedir)
{
	char *buf, *ret;
	struct dentry *d, *tmp;
805
	int buflen = ovl_redirect_max + 1;
M
Miklos Szeredi 已提交
806 807 808 809 810 811 812

	if (samedir) {
		ret = kstrndup(dentry->d_name.name, dentry->d_name.len,
			       GFP_KERNEL);
		goto out;
	}

813
	buf = ret = kmalloc(buflen, GFP_KERNEL);
M
Miklos Szeredi 已提交
814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872
	if (!buf)
		goto out;

	buflen--;
	buf[buflen] = '\0';
	for (d = dget(dentry); !IS_ROOT(d);) {
		const char *name;
		int thislen;

		spin_lock(&d->d_lock);
		name = ovl_dentry_get_redirect(d);
		if (name) {
			thislen = strlen(name);
		} else {
			name = d->d_name.name;
			thislen = d->d_name.len;
		}

		/* If path is too long, fall back to userspace move */
		if (thislen + (name[0] != '/') > buflen) {
			ret = ERR_PTR(-EXDEV);
			spin_unlock(&d->d_lock);
			goto out_put;
		}

		buflen -= thislen;
		memcpy(&buf[buflen], name, thislen);
		tmp = dget_dlock(d->d_parent);
		spin_unlock(&d->d_lock);

		dput(d);
		d = tmp;

		/* Absolute redirect: finished */
		if (buf[buflen] == '/')
			break;
		buflen--;
		buf[buflen] = '/';
	}
	ret = kstrdup(&buf[buflen], GFP_KERNEL);
out_put:
	dput(d);
	kfree(buf);
out:
	return ret ? ret : ERR_PTR(-ENOMEM);
}

static int ovl_set_redirect(struct dentry *dentry, bool samedir)
{
	int err;
	const char *redirect = ovl_dentry_get_redirect(dentry);

	if (redirect && (samedir || redirect[0] == '/'))
		return 0;

	redirect = ovl_get_redirect(dentry, samedir);
	if (IS_ERR(redirect))
		return PTR_ERR(redirect);

873 874 875
	err = ovl_check_setxattr(dentry, ovl_dentry_upper(dentry),
				 OVL_XATTR_REDIRECT,
				 redirect, strlen(redirect), -EXDEV);
M
Miklos Szeredi 已提交
876 877 878 879 880 881
	if (!err) {
		spin_lock(&dentry->d_lock);
		ovl_dentry_set_redirect(dentry, redirect);
		spin_unlock(&dentry->d_lock);
	} else {
		kfree(redirect);
A
Amir Goldstein 已提交
882 883
		pr_warn_ratelimited("overlayfs: failed to set redirect (%i)\n",
				    err);
M
Miklos Szeredi 已提交
884 885 886 887 888 889
		/* Fall back to userspace copy-up */
		err = -EXDEV;
	}
	return err;
}

890 891 892
static int ovl_rename(struct inode *olddir, struct dentry *old,
		      struct inode *newdir, struct dentry *new,
		      unsigned int flags)
M
Miklos Szeredi 已提交
893 894
{
	int err;
895
	bool locked = false;
M
Miklos Szeredi 已提交
896 897 898 899 900 901 902 903 904
	struct dentry *old_upperdir;
	struct dentry *new_upperdir;
	struct dentry *olddentry;
	struct dentry *newdentry;
	struct dentry *trap;
	bool old_opaque;
	bool new_opaque;
	bool cleanup_whiteout = false;
	bool overwrite = !(flags & RENAME_EXCHANGE);
905
	bool is_dir = d_is_dir(old);
906
	bool new_is_dir = d_is_dir(new);
M
Miklos Szeredi 已提交
907
	bool samedir = olddir == newdir;
M
Miklos Szeredi 已提交
908 909
	struct dentry *opaquedir = NULL;
	const struct cred *old_cred = NULL;
910
	LIST_HEAD(list);
M
Miklos Szeredi 已提交
911 912 913 914 915 916 917 918 919

	err = -EINVAL;
	if (flags & ~(RENAME_EXCHANGE | RENAME_NOREPLACE))
		goto out;

	flags &= ~RENAME_NOREPLACE;

	/* Don't copy up directory trees */
	err = -EXDEV;
M
Miklos Szeredi 已提交
920
	if (!ovl_can_move(old))
921
		goto out;
M
Miklos Szeredi 已提交
922
	if (!overwrite && !ovl_can_move(new))
M
Miklos Szeredi 已提交
923 924
		goto out;

925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945
	if (overwrite && new_is_dir && !ovl_pure_upper(new)) {
		err = ovl_check_empty_dir(new, &list);
		if (err)
			goto out;
	}

	if (overwrite) {
		if (ovl_lower_positive(old)) {
			if (!ovl_dentry_is_whiteout(new)) {
				/* Whiteout source */
				flags |= RENAME_WHITEOUT;
			} else {
				/* Switch whiteouts */
				flags |= RENAME_EXCHANGE;
			}
		} else if (is_dir && ovl_dentry_is_whiteout(new)) {
			flags |= RENAME_EXCHANGE;
			cleanup_whiteout = true;
		}
	}

M
Miklos Szeredi 已提交
946 947 948 949 950 951 952 953 954 955 956 957 958 959 960
	err = ovl_want_write(old);
	if (err)
		goto out;

	err = ovl_copy_up(old);
	if (err)
		goto out_drop_write;

	err = ovl_copy_up(new->d_parent);
	if (err)
		goto out_drop_write;
	if (!overwrite) {
		err = ovl_copy_up(new);
		if (err)
			goto out_drop_write;
961 962 963 964
	} else {
		err = ovl_nlink_start(new, &locked);
		if (err)
			goto out_drop_write;
M
Miklos Szeredi 已提交
965 966
	}

967
	old_cred = ovl_override_creds(old->d_sb);
M
Miklos Szeredi 已提交
968

969 970
	if (!list_empty(&list)) {
		opaquedir = ovl_clear_empty(new, &list);
M
Miklos Szeredi 已提交
971 972 973 974 975 976 977 978 979 980
		err = PTR_ERR(opaquedir);
		if (IS_ERR(opaquedir)) {
			opaquedir = NULL;
			goto out_revert_creds;
		}
	}

	old_upperdir = ovl_dentry_upper(old->d_parent);
	new_upperdir = ovl_dentry_upper(new->d_parent);

981 982 983
	if (!samedir) {
		/*
		 * When moving a merge dir or non-dir with copy up origin into
984 985 986
		 * a new parent, we are marking the new parent dir "impure".
		 * When ovl_iterate() iterates an "impure" upper dir, it will
		 * lookup the origin inodes of the entries to fill d_ino.
987
		 */
988
		if (ovl_type_origin(old)) {
989 990 991 992
			err = ovl_set_impure(new->d_parent, new_upperdir);
			if (err)
				goto out_revert_creds;
		}
993
		if (!overwrite && ovl_type_origin(new)) {
994 995 996 997 998 999
			err = ovl_set_impure(old->d_parent, old_upperdir);
			if (err)
				goto out_revert_creds;
		}
	}

M
Miklos Szeredi 已提交
1000 1001
	trap = lock_rename(new_upperdir, old_upperdir);

1002 1003 1004 1005 1006 1007 1008
	olddentry = lookup_one_len(old->d_name.name, old_upperdir,
				   old->d_name.len);
	err = PTR_ERR(olddentry);
	if (IS_ERR(olddentry))
		goto out_unlock;

	err = -ESTALE;
M
Miklos Szeredi 已提交
1009
	if (!ovl_matches_upper(old, olddentry))
1010 1011 1012 1013 1014 1015 1016 1017
		goto out_dput_old;

	newdentry = lookup_one_len(new->d_name.name, new_upperdir,
				   new->d_name.len);
	err = PTR_ERR(newdentry);
	if (IS_ERR(newdentry))
		goto out_dput_old;

1018 1019 1020
	old_opaque = ovl_dentry_is_opaque(old);
	new_opaque = ovl_dentry_is_opaque(new);

1021
	err = -ESTALE;
1022
	if (d_inode(new) && ovl_dentry_upper(new)) {
M
Miklos Szeredi 已提交
1023
		if (opaquedir) {
1024 1025
			if (newdentry != opaquedir)
				goto out_dput;
M
Miklos Szeredi 已提交
1026
		} else {
M
Miklos Szeredi 已提交
1027
			if (!ovl_matches_upper(new, newdentry))
1028
				goto out_dput;
M
Miklos Szeredi 已提交
1029 1030
		}
	} else {
1031 1032 1033
		if (!d_is_negative(newdentry) &&
		    (!new_opaque || !ovl_is_whiteout(newdentry)))
			goto out_dput;
M
Miklos Szeredi 已提交
1034 1035 1036 1037 1038 1039 1040
	}

	if (olddentry == trap)
		goto out_dput;
	if (newdentry == trap)
		goto out_dput;

M
Miklos Szeredi 已提交
1041 1042 1043
	if (WARN_ON(olddentry->d_inode == newdentry->d_inode))
		goto out_dput;

M
Miklos Szeredi 已提交
1044
	err = 0;
M
Miklos Szeredi 已提交
1045
	if (is_dir) {
M
Miklos Szeredi 已提交
1046
		if (ovl_type_merge_or_lower(old))
M
Miklos Szeredi 已提交
1047
			err = ovl_set_redirect(old, samedir);
1048
		else if (!old_opaque && ovl_type_merge(new->d_parent))
1049
			err = ovl_set_opaque_xerr(old, olddentry, -EXDEV);
M
Miklos Szeredi 已提交
1050 1051
		if (err)
			goto out_dput;
M
Miklos Szeredi 已提交
1052
	}
M
Miklos Szeredi 已提交
1053
	if (!overwrite && new_is_dir) {
M
Miklos Szeredi 已提交
1054
		if (ovl_type_merge_or_lower(new))
M
Miklos Szeredi 已提交
1055
			err = ovl_set_redirect(new, samedir);
1056
		else if (!new_opaque && ovl_type_merge(old->d_parent))
1057
			err = ovl_set_opaque_xerr(new, newdentry, -EXDEV);
M
Miklos Szeredi 已提交
1058 1059
		if (err)
			goto out_dput;
M
Miklos Szeredi 已提交
1060 1061
	}

1062 1063 1064
	err = ovl_do_rename(old_upperdir->d_inode, olddentry,
			    new_upperdir->d_inode, newdentry, flags);
	if (err)
M
Miklos Szeredi 已提交
1065 1066 1067 1068 1069
		goto out_dput;

	if (cleanup_whiteout)
		ovl_cleanup(old_upperdir->d_inode, newdentry);

1070 1071 1072 1073 1074 1075 1076
	if (overwrite && d_inode(new)) {
		if (new_is_dir)
			clear_nlink(d_inode(new));
		else
			drop_nlink(d_inode(new));
	}

1077 1078 1079 1080
	ovl_dentry_version_inc(old->d_parent, ovl_type_origin(old) ||
			       (!overwrite && ovl_type_origin(new)));
	ovl_dentry_version_inc(new->d_parent, ovl_type_origin(old) ||
			       (d_inode(new) && ovl_type_origin(new)));
M
Miklos Szeredi 已提交
1081 1082 1083

out_dput:
	dput(newdentry);
1084 1085
out_dput_old:
	dput(olddentry);
M
Miklos Szeredi 已提交
1086 1087 1088
out_unlock:
	unlock_rename(new_upperdir, old_upperdir);
out_revert_creds:
1089
	revert_creds(old_cred);
1090
	ovl_nlink_end(new, locked);
M
Miklos Szeredi 已提交
1091 1092 1093 1094
out_drop_write:
	ovl_drop_write(old);
out:
	dput(opaquedir);
1095
	ovl_cache_free(&list);
M
Miklos Szeredi 已提交
1096 1097 1098 1099 1100 1101 1102 1103 1104
	return err;
}

const struct inode_operations ovl_dir_inode_operations = {
	.lookup		= ovl_lookup,
	.mkdir		= ovl_mkdir,
	.symlink	= ovl_symlink,
	.unlink		= ovl_unlink,
	.rmdir		= ovl_rmdir,
1105
	.rename		= ovl_rename,
M
Miklos Szeredi 已提交
1106 1107 1108 1109 1110
	.link		= ovl_link,
	.setattr	= ovl_setattr,
	.create		= ovl_create,
	.mknod		= ovl_mknod,
	.permission	= ovl_permission,
1111
	.getattr	= ovl_getattr,
M
Miklos Szeredi 已提交
1112
	.listxattr	= ovl_listxattr,
1113
	.get_acl	= ovl_get_acl,
M
Miklos Szeredi 已提交
1114
	.update_time	= ovl_update_time,
M
Miklos Szeredi 已提交
1115
};